1. Field of the Invention
The present invention relates to an environment recognition device and an environment recognition method for recognizing a target object based or a luminance of the target object in a detection area.
2. Description of Related Art
Conventionally, a technique has been known that detects a target object such as an obstacle including a vehicle and a traffic light located in front of a subject vehicle for performing control to avoid collision with the detected target object and to maintain a safe distance between the subject vehicle and the preceding vehicle (for example, Japanese Patent. No 3349060 (Japanese Patent Application Laid-Open (JP-A) No. 10-283461). Further, in such techniques, there is a technique that performs more advanced control. Specifically, it not only specifies a target object simply as a solid object, but further determines whether the detected target object is a preceding vehicle that is running at the same speed as the subject vehicle or a fixed object that does not move. In this case, when the target object is detected by capturing an image of a detection area, it is necessary to extract (cut out) the target object from the captured image before specifying what the target object is.
For example, when the captured image is a color image, there may be a method for grouping pixels having a same luminance (color) and extracting the pixels as a target object.
However, regarding an actual traffic light, an actual road sign, and the like provided on a road, for example, color unevenness can be seen in an image corresponding to a lighting portion of a bulb traffic light, and thus a same target object may not necessarily emit light in a single color. Moreover, a false color may occur based on a Bayer pattern formed in checkered pattern, and therefore, it may be difficult to strictly specify one target object with only one luminance.